Towards an Alarm for Opposition Conflict in a Conjunctive Combination of Belief Functions

Abstract

In the framework of belief functions, information fusion is based on the construction of a unique belief function resulting from the combination of available belief functions induced from several information sources. When sources are reliable and distinct, Smets’ conjunctive rule, which is equivalent to Dempster’s rule of combination without the normalization process, can be considered. This rule offers interesting properties, but in return the empty set is an absorbing element: a series of conjunctive combinations tends to bring a mass equal to 1 to the empty set, making impossible the distinction between a real problem and an effect due to this absorbing effect of the empty set. Then a formalism allowing the preservation of the conflict which reflects the opposition between sources, is introduced in this paper. Based on the normalization process and on distance measures between belief functions, it is tested and compared with classic conjunctive operators on synthetic belief functions.KeywordsInformation FusionCombination RuleBelief FunctionEvidence TheoryApproximate ReasoningThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.